JP2021136023A - Index value predicting device, index value predicting method and program - Google Patents

Abstract

To more accurately, a commercial scene index value of an object plant.SOLUTION: An index value predicting device 10 comprises: a first acquiring part 101 for acquiring data related to previous weather in a specific region; a second acquiring part 102 for acquiring data indicating a previous commercial scene index value of a non object plant being different from an object plant; a third acquiring part 103 for acquiring data indicating a previous commercial scene index value of the object plant; a prediction model storage part 104 for storing a prediction model for predicting a commercial scene index value of the object plant; and a predicting part for inputting the acquired data acquired by the first acquiring part, the acquired data acquired by the second acquiring part, and the acquired data acquired by the third acquiring part to the predicting model, to predict a commercial scene index value of the object plant in a prediction object period.SELECTED DRAWING: Figure 3

Description

The present invention relates to an index value prediction device, an index value prediction method, and a program for predicting a market index value, and more particularly to an index value prediction device, an index value prediction method, and a program for predicting a market index value of a target plant.

The market index value (for example, price, etc.) of a plant such as a vegetable affects the supply and demand of the plant. Producers of vegetables, etc. adjust their production (supply) based on the trends of market index values. For example, when the market price of a certain vegetable becomes high, a person who produces a vegetable of the same kind as the vegetable in a plant factory tries to increase the shipment amount of the vegetable cultivated in the factory.

Techniques for predicting market index values of vegetables and the like have been conventionally developed. For example, according to the technique described in Patent Document 1, it is possible to predict the price of a crop at each of a plurality of candidate times set with respect to the harvest time.

Japanese Unexamined Patent Publication No. 2019-83746

In the index value prediction technology developed so far, including Patent Document 1, for example, the weather at the time of prediction and the past market price of the target plant are used as variables, and the value of these variables is used as the prediction target. The index value of the target crop at the time was predicted.
On the other hand, with regard to the prediction of the market index value of the target plant, higher prediction accuracy is required, and the development of a technique for improving the prediction accuracy of the market index value is expected.

Therefore, the present invention has been made in view of the above circumstances, and an object of the present invention is to predict the market index value of the target plant with higher accuracy.

In order to achieve the above object, the index value prediction device of the present invention is an index value prediction device that predicts the market index value of the target plant, and is the first acquisition unit that acquires data on past weather in a specific area. The second acquisition unit that acquires the past market index value of the non-target plant different from the target plant, the third acquisition unit that acquires the data indicating the past market index value of the target plant, and the target plant. A storage unit that stores a prediction model for predicting the market index value of It is characterized by having a prediction unit that predicts the market index value of the target plant at the target time.

A more preferable configuration of the index value prediction device of the present invention described above will be described. For example, the second acquisition unit acquires data indicating past market index values of a plurality of types of non-target plants, and the second acquisition unit obtains data. Using the acquired data of the above, the calculation unit that calculates the market index value of the target plant of the target plant of the past calculation target time while changing the type of non-target plant, and the target plant of the calculation target time indicated by the acquisition data of the third acquisition unit. The prediction unit has an evaluation unit that evaluates the degree of correlation between the market index value of the above and the market index value of the target plant at the calculation target time calculated by the calculation unit while changing the type of non-target plant. It is preferable to predict the market index value of the target plant at the prediction target time by inputting the data acquired by the second acquisition unit for the non-target plant of the type selected based on the degree into the prediction model.

To further describe a more preferable configuration, a plurality of combinations consisting of two or more types of non-target plants are set, and the calculation unit uses data acquired by the second acquisition unit for each of the two or more types of non-target plants included in each combination. The market index value of the target plant at the calculation target time is calculated by changing the combination of non-target plants, and the evaluation unit calculates the market index value of the target plant at the calculation target time indicated by the acquisition data of the third acquisition unit. And the degree of correlation with the market index value of the target plant at the calculation target time calculated by the calculation unit is evaluated while changing the combination of non-target plants, and the prediction unit is included in the combination selected based on the correlation degree. It is preferable to predict the market index value of the target plant at the prediction target time by inputting the data acquired by the second acquisition unit for each of the two or more types of non-target plants into the prediction model.

To further describe a more preferable configuration, the first acquisition unit acquires data on past weather in a plurality of specific areas, and uses the acquired data of the first acquisition unit to market index of the target plant at the past calculation target time. The calculation unit that calculates the value while changing the specific area, the market index value of the target plant at the calculation target time indicated by the acquisition data of the third acquisition unit, and the market index value of the target plant at the calculation target time calculated by the calculation unit. It has an evaluation unit that evaluates the degree of correlation with and while changing a specific area, and the prediction unit inputs the data acquired by the first acquisition unit for the specific area selected based on the degree of correlation into the prediction model. Therefore, it is advisable to predict the market index value of the target plant at the forecast target time.

To further describe a more preferable configuration, a plurality of combinations consisting of two or more specific areas are set, and the calculation unit uses the data acquired by the first acquisition unit for each of the two or more specific areas included in each combination. The market index value of the target plant at the calculation target time is calculated while changing the combination of specific areas, and the evaluation department uses the market index value of the target plant at the calculation target time indicated by the acquired data of the third acquisition department and the calculation department. The degree of correlation with the market index value of the target plant at the calculated calculation target time is evaluated while changing the combination of specific regions, and the forecasting unit evaluates the degree of correlation with two or more specific regions included in the combination selected based on the degree of correlation. By inputting the data acquired by the first acquisition unit into the prediction model for each of the above, it is preferable to predict the market index value of the target plant at the prediction target time.

More preferably, the second acquisition unit acquires data showing the market index value of the non-target plant in the period before the prediction target time, and the third acquisition unit acquires the data showing the market index value of the non-target plant in the period before the prediction target time. The data showing the market index value of the target plant during the period is acquired, and the prediction unit includes the data showing the market index value of the non-target plant at the specified time among the acquired data of the second acquisition unit and the third acquisition unit. It is preferable to predict the market index value of the target plant at the prediction target time by inputting the data indicating the market index value of the target plant at the specified time and the acquired data of the above into the prediction model.

More preferably, the first acquisition unit may acquire data on past weather in at least one of a specific area where the target plant is produced and a specific area where the wholesale market of the target plant is located. ..

In addition, the forecasting unit may predict a value related to the shipping scale of the target plant as the market price value of the target plant.

In addition, the index value prediction device of the present invention may further have a fourth acquisition unit that acquires data on the domestic infection status of the infectious disease. In this case, the prediction unit inputs the acquisition data of the first acquisition unit, the acquisition data of the second acquisition unit, the acquisition data of the third acquisition unit, and the acquisition data of the fourth acquisition unit into the prediction model, respectively. The market index value of the target plant at the forecast target time may be predicted.

Further, in order to solve the above-mentioned problems, the index value prediction method of the present invention is an index value prediction method for predicting the market index value of the target plant, and the computer acquires data on the past weather in a specific area. The first acquisition process and the data in which the computer acquires the past market index value of the non-target plant different from the target plant, and the second acquisition process and the data in which the computer indicates the past market index value of the target plant. The third acquisition process to acquire the data, the storage process in which the computer stores the prediction model for predicting the market index value of the target plant, the acquisition data in the first acquisition process, and the acquisition data in the second acquisition process by the computer. This is an index value prediction method having a prediction step of predicting the market index value of the target plant at the prediction target time by inputting the data acquired in the third acquisition step into the prediction model, respectively.

Further, in order to solve the above-mentioned problems, the program of the present invention is a program for causing a computer to predict the market index value of the target plant, and includes a first acquisition step of acquiring data on past weather in a specific area. The second acquisition step of acquiring data showing the past market index value of the non-target plant different from the target plant, the third acquisition process of acquiring the data showing the past market index value of the target plant, and the market of the target plant. By inputting the storage process that stores the prediction model that predicts the index value, the acquisition data in the first acquisition process, the acquisition data in the second acquisition process, and the acquisition data in the third acquisition process into the prediction model, respectively. , A program configured to cause a computer to perform a prediction process for predicting market index values of target plants at the time of prediction.

According to the present invention, the market of the target plant at the forecast target time is used by using the past weather in a specific area, the past market index value of the non-target plant different from the target plant, and the past market index value of the target plant. The index value can be predicted accurately.

It is a conceptual diagram of the factory cultivation system including the index value prediction apparatus which concerns on one Embodiment of this invention. It is a figure which shows the hardware equipment which the index value predicting apparatus which concerns on one Embodiment of this invention includes. It is a figure which shows the function of the index value predicting apparatus which concerns on one Embodiment of this invention. It is a figure which shows the flow of the prediction model acquisition phase in the index value prediction flow. It is a figure which shows the flow of the price prediction phase in the index value prediction flow. It is a figure which shows the correlation between the variable about the domestic infection situation of an infectious disease, and the market price (the 1). It is a figure which shows the correlation between the variable about the domestic infection situation of an infectious disease, and the market price (the 2).

An index value prediction device, an index value prediction method, and a program according to an embodiment of the present invention (hereinafter referred to as the present embodiment) will be described.
It should be noted that the embodiments described below are merely examples for facilitating the understanding of the present invention, and do not limit the present invention. That is, the present invention may be modified or improved from the embodiments described below without departing from the spirit of the present invention. Also, of course, the present invention includes an equivalent thereof.

Further, in the present specification, the term "device" refers to one device that independently exerts a specific function, or a plurality of devices that exist in a dispersed manner but cooperate to exert a specific function. Also includes.
Further, in the present specification, the market index value includes the market price, the market transaction volume, and the market sales.
Further, in the present specification, the “market price” is the wholesale price of the target plant in the wholesale market. The "market transaction volume" is the shipment volume or wholesale quantity of the target plant, and the "market sales" is the amount of the value obtained by multiplying the market price by the market transaction volume. In this specification, a case where the target plant is a vegetable (hereinafter, also referred to as a target vegetable) will be described as a specific example. However, the target plant may be a plant other than vegetables, for example, fruits and flowers.
Further, in the present specification, "weather" is a concept indicating the weather, temperature, humidity, atmospheric pressure, precipitation, snowfall, solar radiation and other atmospheric conditions, and the phenomena appearing as a result thereof. ..
Further, as used herein, the term "plant type" means a type represented by any of the order, family, genus and species in the biological classification. That is, plants that are similar to each other but have different species (varieties) correspond to plants of different types.
Further, in the present specification, the “production area” means a place and area where the target vegetables are cultivated. In addition, the "consumption area" means the place and area where the wholesale market for the target vegetables exists.

<< Use of index value prediction device according to this embodiment >>
The index value prediction device according to the present embodiment is a price prediction device (hereinafter referred to as price prediction device 10) that predicts the market price of the target plant. Hereinafter, the use of the price prediction device 10 will be described with reference to FIG. FIG. 1 is a diagram conceptually showing a factory cultivation system S including a price prediction device 10.
The contents described below may also be applied to an index value prediction device for predicting the market transaction volume (shipment volume) or market sales of the target plant.

The price prediction device 10 is used to predict the market price of the target vegetable at the target time of prediction. The forecast target time is any time in the future, for example, one day or several days later, one week or several weeks later, or one month or several months later. Since the market price generally fluctuates on an hourly, daily, weekly, monthly or yearly basis, the predicted market price for the forecasted period may be the set price at any time during that period. It may be a statistically processed value such as an average value and a maximum value, a minimum value, a median value, an integrated value, a standard deviation, or a mode value that considers or does not consider the market holidays at the time.

The forecast result of the market price (estimated price) is reflected in the cultivation conditions of the target vegetables.
More specifically, in the present embodiment, the target vegetables are cultivated in the field and also in the plant factory F shown in FIG. 1, for example, the target vegetables are cultivated by a hydroponic cultivation method. Specifically, a plurality of target vegetable strains (individuals) are placed on each stage of a multi-stage shelf, and while immersing each root in a nutrient solution, artificial light and / or sunlight is directed toward each leaf and stem. Irradiate. Each strain grows to harvest size and is then harvested and shipped.

The plant factory F is provided with a factory cultivation system S, and the cultivation conditions of the target vegetables in the plant factory F are adjusted by this system. At this time, the estimated price of the target vegetable is input to the control device (not shown) included in the factory cultivation system S.

The control device sets the components in the nutrient solution, the concentration of each component, the intensity of the irradiation light, the temperature and humidity in the shelf, the carbon dioxide concentration, and the like according to the predicted price. Thereby, the growth rate of the target vegetable in the plant factory F is controlled. As a result, the person who produces the target vegetable at the plant factory F can determine the sales of the target vegetable from the predicted price and adjust the shipment amount from the plant factory F to the amount according to the predicted price.

The target vegetables are not limited to the vegetables cultivated in the plant factory F, and the types of target vegetables, cultivation methods, cultivated areas (production areas), wholesale areas (consumption areas), and cultivation The timing is not particularly limited.

Further, in the present embodiment, the forecast result (estimated price) of the market price of the target vegetable is reflected in the cultivation conditions of the target vegetable in the plant factory F, but how to use the predicted price is particularly important. Not limited.

Incidentally, examples of vegetables to which the present invention can be applied include the following vegetables.
Nameko mushrooms, matsutake mushrooms, nameko mushrooms, enoki mushrooms, shimeji mushrooms, mushrooms, maitake mushrooms, king trumpet mushrooms, and other mushrooms;
Bracken, mountain udo, tara sprout, and other wild plants;
Ume, Yuzu, Daidai, Sudachi, Kabosu, Ginkgo, and other fruit trees;
Cucumbers, pumpkins, zucchini, eggplants, sardines, bein sardines, eggplants, tomatoes, cherry tomatoes, peppers, jumbo peppers, paprika, sardines, corn, okra, corn, sardines, sardines, and other fruit vegetables ;
Wasabi, root ginger, leaf ginger, sardine, rice, sardine, shiso, aunt, hojiso, myoga, myogatake, herbs, bofu, morokuri, fukinoto, edible chrysanthemum, baby leaf, and others Spicy ginger;
Radish, turnip, carrot, burdock, bamboo shoot, lotus root, arrowhead, radish, and other root vegetables;
Make-in, Baron, Baron, Kansho, Satoimo, Celebes, Kyoimo, Hoopoe, Nagaimo, Yamatoimo, and other potatoes;
Onions, garlic, scallions, shallot, yurine, pekoros, allylets, and other souvenirs and vegetables;
Green beans, pods, pea, pea, broad beans, edamame, and other legumes;
Lettuce, Sunny Lettuce, Green Leaf Lettuce, Cabbage, Green Ball, Hakusai, Santosai, Mizuna, Tsukina, Komatsuna, Tsumamina, Spinach, Green Onion, Konegi, Divided, Asatsuki, Wipe, Udo, Root Mitsuba , Cut mitsuba, thread mitsuba, shungiku, seri, tomorrow, nira, cerule, asparagus, potash flower, broccoli, salad, parsley, garlic sprout, bok choy, lettuce, nanohana, red cabbage, lettuce, cresson, endive , Moroheiya, lettuce, green onions, and other leaf stem vegetables.

<< Configuration of price forecaster >>
Next, the configuration of the price forecasting device 10 will be described with reference to FIGS. 2 and 3. FIG. 2 is a diagram showing a hardware device included in the price prediction device 10. FIG. 3 is a diagram showing the functions of the price prediction device 10.
The price prediction device 10 is composed of a computer. The computer constituting the price prediction device 10 corresponds to, for example, a terminal (for example, a PC, a tablet-type terminal, a smartphone, etc.) used by the producer of the target vegetable. As shown in FIG. 2, the computer constituting the price prediction device 10 includes a processor 11, a memory 12, a communication interface 13, a storage 14, an input device 15, and an output device 16.

The processor 11 includes a CPU (Central Processing Unit), an MPU (Micro-Processing Unit), an MCU (Micro Controller Unit), a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), a TPU (Tensor Processing Unit), or an ASIC (Application). It is preferable to be composed of Specific Integrated Circuit) or the like. The processor 11 may be an analog arithmetic unit or an arithmetic unit using a physical state peculiar to quantum mechanics.
The memory 12 may be composed of a semiconductor memory such as a ROM (Read Only Memory) and a RAM (Random Access Memory).
The communication interface 13 may be composed of, for example, a network interface card, a communication interface board, or the like. By the way, the standard of data communication by the communication interface 13 is not particularly limited, and communication by wireless LAN based on Wi-fi (registered trademark), communication by 3G, 4G or 5G mobile communication system, or LTE. Communication based on (Long Term Evolution) can be mentioned.
The storage 14 includes a flash memory, an HDD (Hard Disc Drive), an SSD (Solid State Drive), an FD (Flexible Disc), an MO disc (Magneto-Optical disc), a CD (Compact Disc), a DVD (Digital Versatile Disc), and an SD. It may be composed of a card (Secure Digital card), a USB memory (Universal Serial Bus memory), or the like.
The input device 15 may be composed of, for example, a keyboard, a mouse, a touch panel, or the like. The output device 16 may be composed of, for example, a display, a speaker, or the like.

Further, a program for price prediction is installed as software on the computer constituting the price prediction device 10. This program corresponds to the program of the present invention and is read and executed by the processor 11. By executing this program, the computer functions as the price prediction device 10 and performs various data processing for predicting the market price of the target vegetable.
The above program may be acquired by reading it from a computer-readable recording medium, or may be acquired by receiving (downloading) it via the Internet, an intranet, or the like.

Further, the computer constituting the price prediction device 10 communicates with another computer such as a server computer and acquires data through a communication network such as the Internet and a communication interface 13. Other computers include, for example, a computer that distributes data related to the weather in each region, a computer that distributes data indicating the market price of vegetables, and the like.

The configuration of the computer constituting the price prediction device 10 will be newly described from the functional aspect. As shown in FIG. 3, the computers constituting the price prediction device 10 include a first acquisition unit 101, a second acquisition unit 102, a third acquisition unit 103, a storage unit 104, a prediction unit 105, a calculation unit 106, and an evaluation unit. It has 107 as a functional unit.
The storage unit 104 is composed of the memory 12 or the storage 14 described above. The functional units other than the storage unit 104 are realized by the cooperation of the above-mentioned hardware device included in the computer constituting the price prediction device 10 and the above-mentioned price prediction program.
Hereinafter, each of the above-mentioned functional units will be described.

(1st acquisition department)
The first acquisition unit 101 acquires data on past weather in a specific area by communicating with another computer via the Internet or the like.
In the present embodiment, the "specific area" is an area divided according to administrative divisions, and in Japan, it is a prefecture or a municipality. However, the present invention is not limited to this, and the area may be divided into smaller units than the municipality.

In the present embodiment, the data acquired by the first acquisition unit 101 is, for example, the average temperature (° C.), the average humidity (%), the number of foggy days (days), the average pressure (hPa), and the maximum value of 10-minute precipitation ( mm), total amount of snowfall (cm), number of days (days) when the daily average cloud amount is less than 1.5, number of days (days) when the daily precipitation is 0.0 mm or more, and average total solar radiation (mm) The data may indicate MJ / m2).
The value indicated by the acquired data of the first acquisition unit 101 will be referred to as a "weather value" below for convenience of explanation.

In the present embodiment, the first acquisition unit 101 acquires data on past weather in a plurality of specific areas, and for example, acquires data on weather in various parts of Japan during a predetermined period. Here, the predetermined period is a period preset by the user of the price prediction device 10, and the length of the predetermined period is arbitrarily determined. In the following description, the predetermined period is determined on a monthly basis, and for example, it is assumed that the period is from n months before the current time (n is a natural number) to the present time.
The meteorological value indicated by the acquired data of the first acquisition unit 101 may be a representative value of each month in the above-mentioned predetermined period, for example, an average value, a maximum value, a minimum value, a median value, an integrated value, a standard deviation, and the like. Alternatively, it may be a statistically processed value such as the mode value.

In addition, it is preferable that the specific area includes at least one of the production area and the consumption area of the target vegetable. That is, the first acquisition unit 101 may acquire data on past weather in at least one of the specific area where the target vegetable is produced and the specific area where the target vegetable is consumed.

(2nd acquisition department)
The second acquisition unit 102 acquires data indicating the past market price of the non-target vegetable different from the target vegetable by communicating with another computer via the Internet or the like. The non-target vegetable corresponds to the non-target plant and is a kind of vegetable different from the target vegetable.
The non-target vegetables are preferably selected in relation to the target vegetables. For example, if the target vegetable is a vegetable that grows on the ground (specifically, a vegetable in which the main edible portion appears on the ground), it is better to select a non-target vegetable that also grows on the ground.

Further, in the present embodiment, the second acquisition unit 102 acquires data indicating the past market prices of a plurality of types of non-target vegetables, for example, for each type of non-target vegetables for a predetermined period (n months before the present time). Obtain data showing the monthly average price of each month for the period up to this point.

(Third acquisition department)
The third acquisition unit 103 acquires data indicating the past market price of the target vegetable by communicating with another computer via the Internet or the like. In the present embodiment, the third acquisition unit 103 acquires data indicating the monthly average price of each month in a predetermined period (the period from n months before the present time to the present time) for the target vegetables.

(Memory)
The storage unit 104 stores the acquired data of each of the first acquisition unit 101, the second acquisition unit 102, and the third acquisition unit 103, and also stores a prediction model for predicting the market price of the target vegetable.
In the present embodiment, the storage unit 104 is composed of storage devices (specifically, the memory 12 and the storage 14) included in the computer constituting the price prediction device 10, but is not limited thereto. For example, various data may be stored in an external server (specifically, a server for a cloud service or the like) that is communicably connected to the price prediction device 10.

（１） The prediction model is a function (approximate formula) or a mathematical model for predicting the market price of the target vegetable, and is the following prediction formula (1) in the present embodiment.

(1)

In the above prediction formula (1), Pm on the left side is a predicted value of the market price of the target vegetable at the prediction target time, that is, the predicted price. P (si), Q (gc, tj) and R (hd, uk) on the right side are variables, that is, input values for price prediction.
P (si) is the market price of the target vegetable in the i-th (i is a natural number of 1 to x) designated month si in the period before the forecast target time, strictly speaking, the monthly average price. Here, the designated month si is a month designated in advance for determining the value used for price prediction.
Q (gc, tj) is the market price of non-target vegetables of the c-th type (c is a natural number of 1 to m), and is the j-th (j is 1 to y) in the period before the forecast target time. The market price in the designated month tj, strictly speaking, the monthly average price.
R (hd, uk) is the meteorological value in the dth (d is a natural number of 1 to n) special area hd, and the meteorological value in the kth designated month uk in the period before the forecast target time, strictly speaking. Is a representative value such as the average value for the month.
The symbols m, n, x, y, and z above represent arbitrary natural numbers.

Further, the parameters ai, aj, ak and b in the prediction formula (1) are determined by the least squares method. That is, the difference between the forecast result of the market price and the actual market price is calculated, and each of the above parameters is determined so that the sum of squares of the difference is minimized.

Then, by inputting each value indicated by the acquired data of the first acquisition unit 101, the second acquisition unit 102, and the third acquisition unit 103 into the prediction formula (1), the market price of the target vegetable at the prediction target time is obtained. Can be predicted (estimated).

In addition, on the right side of the above-mentioned prediction formula (1), not only the respective terms of P (si), Q (gc, tj) and R (hd, uk) are added, but also these may be multiplied. Specifically, P (si) * Q (gc, tj) and Q (gc, tj) * R (hd, uk), that is, a term indicating the product of the market price and the meteorological data may be included. .. Further, the prediction formula (1) may include not only multiplication by multiplying the above terms but also operations such as division or exponentiation.

Further, the prediction model is not limited to the approximation formula such as the prediction formula (1) above or the regression formula by multivariate analysis, and may be another model. For example, machine learning is performed using the data set including the acquired data of the first acquisition unit 101, the second acquisition unit 102, and the third acquisition unit 103 as training data, and the mathematical model obtained as a result is predicted. It may be a model. This prediction model uses the acquired data of each of the first acquisition unit 101, the second acquisition unit 102, and the third acquisition unit 103 as input data, and outputs the market price of the target vegetable at the prediction target time.

When the prediction model is acquired by machine learning, any type of learning may be used: supervised learning, unsupervised learning, and reinforcement learning. Further, the machine learning algorithm may be an error backpropagation method, a gradient descent method, a simulated annealing method, a quantum annealing method or other methods. In addition, the mathematical models constructed by machine learning are neural networks, convolutional neural networks, recurrent neural networks, attentions, transformers, hostile generation networks, deep learning neural networks, Boltzmann machines, matrix factorization (MF), and fax. Learning Machine (FM), M-way Factorization Machine (M-way FM), Field Recognition Factorization Machine (FFM), Field Recognition Neural Factorization Machine (FNFM), Support Vector It may be a machine, a basic network, a decision tree, a random forest, or any other model. When the above-mentioned fact-ization machine (MF, FM, M-way FM, FFM, FNFM) is adopted, the prediction formula is defined by the model equation of factorization. In this case, the term of the combination feature quantity (alternate term) included in the prediction formula may be in the form of Quadratic Unconstrained Binary Optimization (QUADO), or converted to the Ising model. It may be calculated using quantum annealing.

In addition, each parameter in the prediction model, including ai, aj, ak and b in the prediction formula (1), may be fixed once determined, or the prediction target time and the type of target vegetable, etc. It may be changed or updated as appropriate according to the above.
Further, when constructing a prediction model with each of the factification machines (MF, FM, M-way FM, FFM, FNFM) listed above, ai, aj, ak and b are vectors, vector norms. Vector inner product, vector outer product, vector Hadamard product, matrix, matrix norm, matrix inner product, matrix outer product, matrix Hadamard product, tensor, tensor norm, tensor inner product, tensor outer product, tensor adamar product And so on.

Further, with respect to the acquired data of the first acquisition unit 101, the second acquisition unit 102, and the third acquisition unit 103, new data is automatically acquired and updated periodically (for example, every week). good. In this case, it is preferable to update each parameter in the prediction model with the update of the acquired data, and specifically, to re-execute the training for constructing the prediction model.

Further, in the present embodiment, the price prediction device 10 derives the prediction formula (1) which is a prediction model, but the present invention is not limited to this. For example, after the prediction formula (1) is derived by another computer, the computer provides the price prediction device 10 with a calculation program including the prediction formula (1), and the price prediction device 10 uses the program. It may be in the form.

(Prediction department)
The prediction unit 105 inputs the acquisition data of the first acquisition unit 101, the acquisition data of the second acquisition unit 102, and the acquisition data of the third acquisition unit 103 into the prediction model, respectively, and the market price of the target vegetable at the prediction target time. Predict. In the present embodiment, the prediction unit 105 reads the prediction formula (1) stored in the storage unit 104 and substitutes the value corresponding to the above-mentioned variable into the prediction formula (1) to obtain the target vegetable at the prediction target time. Predict the market price of.

In the present embodiment, the prediction unit 105 uses the past market prices of two or more types of non-target vegetables when inputting the past market prices of non-target vegetables as the variable Q into the prediction formula (1). be able to. Similarly, when the prediction unit 105 inputs the past weather value in the specific area as the variable R into the prediction formula (1), the prediction unit 105 can use the past weather value in each of two or more specific areas.

(Calculation part)
The calculation unit 106 uses the acquired data of the first acquisition unit 101, the second acquisition unit 102, and the third acquisition unit 103 to market the target vegetables at a predetermined time in the past (hereinafter referred to as the calculation target time). Calculate the price. The calculation target time is an arbitrary time in the past, and may be, for example, last month.

Using the acquired data of the second acquisition unit 102, the calculation unit 106 calculates the market price of the target vegetables at the calculation target time while changing the type of the non-target vegetables.
More specifically, in the present embodiment, a plurality of combinations consisting of two or more kinds of non-target vegetables are set.
The calculation unit 106 uses the data acquired by the second acquisition unit 102 for each of the two or more types of non-target vegetables included in each combination, and calculates the market price of the target vegetables at the calculation target time while changing the combination. The data used at this time is data showing the market price of non-target vegetables in each month from e months before the calculation target time (e is a natural number) to 1 month before the acquisition target time of the second acquisition unit 102. ..
The calculation unit 106 substitutes the value indicated by the above data (that is, the market price of two or more kinds of non-target vegetables) into a predetermined calculation formula for each combination. This calculation formula is a formula prepared for calculating the market price of the target vegetables at the calculation target time from the past market prices of two or more types of non-target vegetables, and is, for example, an approximate formula of a linear linear type.
According to the above procedure, the calculation unit 106 calculates the market price of the target vegetable at the calculation target time for each combination of the non-target vegetables.

In the present embodiment, a combination consisting of two or more kinds of non-target vegetables is set, but the number of types of non-target vegetables included in one combination may be one.

In addition, the calculation unit 106 calculates the market price of the target vegetable at the calculation target time while changing the specific area by using the acquisition data of the first acquisition unit 101.
More specifically, in the present embodiment, a plurality of combinations of two or more specific areas are set.
The calculation unit 106 uses the data acquired by the first acquisition unit 101 for each of the two or more specific areas included in each combination, and calculates the market price of the target vegetable at the calculation target time while changing the combination. The data used at this time is data showing the weather value in a specific area in each month from f months before the calculation target time (f is a natural number) to one month before the acquisition data of the first acquisition unit 101.
For each combination, the calculation unit 106 substitutes the value indicated by the above data (that is, the weather value in two or more specific areas) into a predetermined calculation formula. This calculation formula is a formula prepared for calculating the market price of the target vegetable at the calculation target time from the past weather values in two or more specific areas, and is, for example, a linear linear type approximation formula.
According to the above procedure, the calculation unit 106 calculates the market price of the target vegetable at the calculation target time for each combination of the specific regions.

In the present embodiment, a combination consisting of two or more specific areas is set, but the number of specific areas included in one combination may be one.

(Evaluation department)
The evaluation unit 107 includes the market price of the target vegetable at the calculation target time indicated by the acquisition data of the third acquisition unit 103 (that is, the actual market price) and the market price of the target vegetable at the calculation target time calculated by the calculation unit 106. Evaluate the degree of correlation of. Here, the degree of correlation is an index showing the correlation between the actual market price and the market price calculated by the calculation unit 106, and the correlation coefficient is used as the degree of correlation in the present embodiment. The method for obtaining the correlation coefficient is known, and the description thereof will be omitted.

When the calculation unit 106 calculates the market price of the target vegetables at the calculation target time while changing the types of the non-target vegetables (strictly speaking, a combination consisting of two or more types of non-target vegetables), the evaluation unit 107 does not. The above correlation coefficient is evaluated while changing the type of target vegetables (combination of non-target vegetables).

Further, when the calculation unit 106 calculates the market price of the target vegetable at the calculation target time while changing the specific area (strictly speaking, a combination consisting of two or more specific areas), the evaluation unit 107 calculates the specific area (specific). Calculate the above correlation coefficient while changing the combination of regions).

The correlation coefficient evaluated by the evaluation unit 107 is reflected in the price prediction by the prediction unit 105.
Specifically, the prediction unit 105 selects the type of non-target vegetables based on the correlation coefficient calculated by the evaluation unit 107, and strictly speaking, any one of a plurality of combinations of non-target vegetables set. Select a combination of.
Then, the prediction unit 105 uses the data acquired by the second acquisition unit 102 for the selected types of non-target vegetables (that is, two or more types of non-target vegetables included in the selected combination), and predicts the prediction target time. Forecast the market price of the target vegetables. That is, the prediction unit 105 inputs the past market prices of the two or more selected non-target vegetables into the prediction formula (1), and predicts the market price of the target vegetables at the prediction target time.

Further, the prediction unit 105 selects a specific area based on the correlation coefficient calculated by the evaluation unit 107, and strictly speaking, selects one of a plurality of combinations of the set specific areas.
Then, the prediction unit 105 uses the data acquired by the first acquisition unit 101 for the selected specific area (that is, two or more specific areas included in the selected combination) to market the target vegetables at the prediction target time. Predict the price. That is, the prediction unit 105 inputs the past weather values of two or more selected specific areas into the prediction formula (1), and predicts the market price of the target vegetable at the prediction target time.

<< About price forecast flow >>
Next, the price prediction flow by the price prediction device 10 will be described with reference to FIGS. 4 and 5. 4 and 5 are explanatory diagrams of the price forecast flow.
In the price forecast flow, the price forecast method of the present invention is adopted. That is, each step in the price forecasting flow shown in FIGS. 4 and 5 corresponds to a component of the price forecasting method of the present invention.

The price forecast flow is divided into a forecast model acquisition phase shown in FIG. 4 and a price forecast phase shown in FIG.
Explaining the flow of the prediction model acquisition phase with reference to FIG. 4, first, the computer forming the price prediction device 10 (hereinafter, simply referred to as a computer) determines the weather in the past specific period for each of the plurality of specific areas. Acquire data about (S001).
Here, the past specific period is determined based on the calculation target time of the market price in step S006 described later, and is, for example, a period from m months before the calculation target time to the calculation target time. Further, the data acquired in step S001 is data indicating the average value of the meteorological values of each month in the above-mentioned specific period.

In addition, the computer acquires data indicating the market price in the past specific period for each of the plurality of types of non-target vegetables (S002). The data acquired in step S002 is data showing the monthly average price of non-target vegetables for each month from m months before the calculation target time to the calculation target time.
In addition, the computer uses data showing the market price of the target vegetables in the past specific period, specifically, the monthly average price of the target market price for each month from m months before the calculation target time to the calculation target time. The indicated data is acquired (S003).

Next, the computer calculates the market price of the target vegetable at the calculation target time using the acquired data in S001 to S003 (S004).
In this step S004, the computer uses the weather values of each month from m months to 1 month before the calculation target time among the past weather values in the specific area indicated by the data acquired in step S001 to be calculated. Calculate the market price of the target vegetables for the season.
At this time, the computer sets a plurality of combinations of specific areas, and while changing the combinations, calculates the market price of the target vegetables at the calculation target time using the weather values of each of the two or more specific areas included in each combination. do.

Further, in step S004, the computer uses the market price of each month from m months to 1 month before the calculation target time among the past market prices of non-target vegetables indicated by the data acquired in step S002. Calculate the market price of the target vegetables at the target time.
At this time, the computer sets a plurality of combinations of non-target vegetables, and while changing the combinations, uses the past market prices of each of the two or more types of non-target vegetables included in each combination to calculate the target vegetables at the target time. Calculate the market price of.

Further, in step S004, the computer calculates using the market price of each month from m months to 1 month before the calculation target time among the past market prices of the target vegetables indicated by the data acquired in step S003. Calculate the market price of the target vegetables at the target time.
At this time, the computer specifies one or more months in the period from m months to one month before the calculation target time, and while changing the designated month, uses the market price of the designated month to calculate the calculation target. Calculate the market price of the target vegetables for the season.

After that, the computer evaluates the correlation coefficient between the actual price indicated by the acquired data in step S003 and the price calculated in step S004 for the market price of the target vegetable at the calculation target time (S005).
Specifically, the computer evaluates the correlation coefficient between the price calculated while changing the combination of non-target vegetables and the actual price while changing the combination of non-target vegetables.
In addition, the computer evaluates the correlation coefficient between the price calculated while changing the combination of specific regions and the actual price while changing the combination of specific regions.
Furthermore, the computer changes the correlation coefficient between the calculated price and the actual price while changing the specified month (designated month) in the period from m months to 1 month before the calculation target time, and changes the specified month. Evaluate while.

Next, the computer derives the prediction formula (1) as a prediction model (S006). When deriving the prediction formula (1), the computer sets the variables P, Q, and R in the prediction formula (1) based on the correlation coefficient evaluated in step S005.
Specifically, the designated month with the highest correlation coefficient is selected from the designated months in the period from n months to 1 month before the calculation target time, and the price of the target vegetables in the selected designated month is selected. Is set as the variable P.
In addition, among the combinations of non-target vegetables, two or more types of non-target vegetables included in the combination having the highest correlation coefficient are selected, and the past market prices of the selected two or more types of non-target vegetables are used as the variable Q. Set.
Further, among the combinations of specific areas, two or more specific areas included in the combination having the highest correlation coefficient are selected, and the past weather values of the selected two or more specific areas are set as the variable R.

After that, the computer uses the values of the set variables and the actual market price of the target vegetable indicated by the acquired data in step S003 to set the parameters ai, aj, ak and b in the prediction formula (1). decide. As a result, the prediction formula (1) is derived.

The computer stores the prediction formula (1) determined through the above procedure in a memory or the like in the computer (S007). This step S007 corresponds to the storage step. When step S007 ends, the prediction model acquisition phase ends.
In addition, the forecast model acquisition phase is repeatedly carried out by changing the item of the target vegetable, the location of the wholesale market of the target vegetable, the region corresponding to the specific region, the forecast target time, and the like. As a result, the forecast formula (1) is derived for each item, each market location, each specific region, and each forecast target time.

Next, the flow of the price forecasting phase will be described with reference to FIG.
Prior to the implementation of the price forecasting phase, a computer user (eg, the producer of the target vegetable) launches a program for price forecasting. When the program is started, the input screen is displayed on the computer display. The user inputs the information necessary for price forecasting through the input screen. The input information includes the item of the target vegetable, the location of the wholesale market, the specific area, the forecast target time, and the like. The input information may include other information related to price forecasting (eg, information about the user, etc.).

In the price forecasting phase, the computer first acquires data on past weather in two or more specific regions selected in step S006 of the forecasting model acquisition phase (S011). This step S011 corresponds to the first acquisition step. The data acquired in step S011 is data showing the average value of the meteorological values in each specific area for each month from n months to 1 month before the forecast target time.

In addition, the computer acquires data indicating the past market prices of the two or more types of non-target vegetables selected in step S006 of the prediction model acquisition phase (S012). This step S012 corresponds to the second acquisition step. The data acquired in step S012 is data showing the monthly average price of each non-target vegetable for each month from n months to 1 month before the forecast target time.

Further, the computer acquires data indicating the market price of the target vegetable in the past designated month selected in step S006 of the prediction model acquisition phase (S013). This step S013 corresponds to the third acquisition step. The data acquired in step S013 is data indicating the monthly average price of the target vegetables in the designated month.

After that, the computer uses each of the acquired data in steps S011 to S013, and specifically inputs the value indicated by each acquired data into the prediction formula (1) to predict the market price of the target vegetable at the prediction target time. (S014). This step S014 corresponds to the prediction step. As the prediction formula (1) used in step S014, a prediction formula corresponding to user input information (specifically, target vegetable item, market location, specific area, and prediction target time) is used. It is specified.

In step S014, the past weather values of two or more specific areas included in the combination selected based on the correlation coefficient, and the past of two or more types of non-target vegetables included in the combination selected based on the correlation coefficient. The market price of the target vegetables in the designated month selected based on the correlation coefficient and the market price of the target vegetables are used.

After performing step S014, the computer outputs the predicted market price (predicted price) of the target vegetable through an appropriate means such as display (S015). At this time, information related to the expected price or information derived based on the expected price may be output together. For example, a numerical value indicating the reliability (accuracy) of the predicted price can be displayed, a graph showing the transition of the market price so far can be displayed, the cultivation conditions of the target vegetables according to the predicted price, and the appropriate production amount of the target vegetables can be displayed. Alternatively, a rational production plan of the target vegetable may be displayed.
When the series of steps up to the above is completed, the price forecasting phase ends.

<< About the effectiveness of this embodiment >>
In the present embodiment, when predicting the market price of the target vegetable at the forecast target time, the past weather of the specific area, the past market price of the non-target vegetable, and the past market price of the target vegetable are input values (variables). ). This makes it possible to accurately predict the market price of the target vegetables.
Hereinafter, the effectiveness of the present embodiment will be described from the viewpoint of the correlation between the market price of the target vegetable predicted using the above input values and the actual market price.

（２）
上記の式（２）のうち、Ｐｉは、予測対象月ｍのｉカ月前（ｉは１〜ｘの自然数）のレタスの市場価格であり、ａｉ及びｂは、最小二乗法により求められるパラメータである。 Market prices for fresh foods, including vegetables, generally tend to correlate with earlier market prices. For example, the market price of lettuce in one month can be roughly predicted from the price in the previous months.
Here, for lettuce traded in a certain wholesale market, the market price from January 2003 to September 2019 was predicted using the market price from January 2002 to September 2019. Specifically, the forecast target month is m, the forecast value of the market price of the forecast target month m is Pm, and the forecast value Pm is set to the period before the forecast target month m by the following formula (2). Forecast using market prices.

(2)
In the above formula (2), Pi is the market price of lettuce i months before the forecast target month m (i is a natural number of 1 to x), and ai and b are parameters obtained by the least squares method. be.

When the correlation coefficient between the market price Pm of the forecast target month m predicted by the above formula (2) and the actual price of the same month was obtained, the results shown in Table 1 below were obtained.

As can be seen from Table 1, when x = 1, that is, when using only the market price one month before the forecast target month m, the correlation coefficient with the actual price is 0.6. Below. It was also found that the larger x is, the higher the correlation coefficient is.

Further, when predicting the market price of the target vegetable in the target period, it is preferable to use the market price of the month before the forecast target period, and it is more preferable to further add the market price of the other months. In particular, it is preferable to add the market price of any month during the period from 12 months to 2 months before the forecast target time, and above all, 2 months, 5 months, 6 months or 12 months before the forecast target time. The market price of a month ago is suitable.

（３）
上記の式（３）において、Ｑｊ（ｇ）は、種類ｇの野菜の、予測対象月ｍからｊカ月前（ｊは１〜ｘの自然数）の市場価格であり、それ以外の変数は、式（２）と同様である。また、パラメータａｉ、ａｊ及びｂは、最小二乗法により求められるパラメータである。なお、上記の式（３）のｘは、１から＋１ずつ増やして最大で６とした。 Next, the market price Pm of lettuce in the forecast target month m was predicted by the following formula (3) using the past market prices of vegetables other than lettuce.

(3)
In the above formula (3), Qj (g) is the market price of the vegetable of type g j months before the forecast target month m (j is a natural number of 1 to x), and the other variables are the formula. It is the same as (2). Further, the parameters ai, aj and b are parameters obtained by the least squares method. The x in the above equation (3) was increased by +1 from 1 to a maximum of 6.

Regarding the market price of lettuce in the forecast target month m, the correlation coefficient between the price Pm predicted while changing the type g using the past market price Pi of lettuce and the market price Qj of vegetables of type g and the actual price It is shown in Table 3. Table 3 shows the correlation coefficient between the market price predicted using only the past market price Pi of lettuce and the actual price as comparative values.

As can be seen from Table 3, a higher correlation can be obtained when using each market price of other vegetables and lettuce than when using only the market price of lettuce.
In predicting the market price of lettuce, it is better to use the market price of vegetables that grow on the ground (cucumber, eggplant, and spinach) to obtain the market price of vegetables that grow underground (baron and onion). Correlation is higher than when used. It is considered that this is because lettuce is easily affected by the weather, while vegetables growing underground are not easily affected by the weather.

Based on the above results, it is preferable to use the past market price of the non-target vegetable together with the past market price of the target vegetable when forecasting the market price of the target vegetable at the forecast target time. As the market price of non-target vegetables used for forecasting, the market price of the month before the forecast target period is preferable, and it is more preferable to add the market price of other months.
Further, as the non-target vegetable, it is preferable to select a suitable type of vegetable according to the target vegetable. For example, when the target vegetable is a vegetable that grows on the ground, the non-target vegetable is also a vegetable that grows on the ground. , It is better to predict the market price of the target vegetable using the market price of the non-target vegetable.

（４）
上記の式（４）において、Ｒｋ（ｈ）は、予測対象月ｍからｋカ月前（ｋは１〜ｘの自然数）の地域ｈの気象値であり、それ以外の変数は、式（２）と同様である。また、パラメータａｉ、ａｋ及びｂは、最小二乗法により求められるパラメータである。なお、上記の式（４）中のｘは、５とした。 Next, the market price Pm of lettuce in the target month m was predicted by the following formula (4) using the past weather values (specifically, the monthly average temperature) in various parts of Japan.

(4)
In the above equation (4), Rk (h) is the meteorological value of the region h k months before the forecast target month m (k is a natural number of 1 to x), and the other variables are the equation (2). Is similar to. Further, the parameters ai, ak and b are parameters obtained by the least squares method. In addition, x in the above formula (4) was set to 5.

Table 4 shows the correlation coefficient between the price Pm predicted while changing the region h using the past market price Pi of lettuce and the weather value Rk of the region h and the actual price for the market price of lettuce in the forecast target month. show. By the way, region h is the prefectural capital of all prefectures in Japan.
The correlation coefficient was evaluated for all prefectures and for each region h, but Table 4 shows only the correlation coefficient for some prefectures.

As shown in Table 4, there is no great difference in the correlation coefficient between the regions h. From this, it can be seen that, for example, even if the meteorological value of the lettuce producing area is used, the correlation coefficient does not always increase.
On the other hand, it is considered that the market price of the target vegetable can be predicted with higher accuracy by using the meteorological value of the region h having a high correlation coefficient.

Therefore, a plurality of combinations of two regions h are set, and the lettuce market of the forecast target month m is used by using the past weather values in the two regions h (first region and second region) included in each combination. The price Ms was predicted, and the correlation coefficient between the predicted price and the actual price was obtained. The results are shown in Table 5.

As can be seen from Table 5, there are combinations of regions h with higher correlation coefficients, for example, combinations of regions in the Chubu region and regions in the Kansai region, and regions in the Shikoku or Kyushu area and the Honshu area. In combination with the region of, the correlation coefficient is high. By using the weather values of the regions included in the combination with such a high correlation coefficient, the market price of lettuce can be predicted with higher accuracy.
If the type of the target vegetable changes, the combination of regions h having a high correlation coefficient may change. Therefore, it is preferable to specify the combination of regions having a high correlation coefficient for each type of target vegetable.

(Specific example of prediction formula)
A specific example of the lettuce market price forecast formula derived based on the contents explained above will be described.
In the prediction formula according to the specific example, for example, the input values (variables) shown in Table 6 are used. The correlation coefficient between the lettuce market price Pm of the forecast target month m predicted using the input values (variables) shown in Table 6 and the actual market price is 0.887.
By the way, each of the 13 meteorological data shown in Table 6 is, for example, data of a region selected in descending order of lettuce shipment amount.

The prediction formula according to the specific example is within the scope of the present invention, and the correlation coefficient between the predicted value by the above prediction formula and the actual price is high for the market price of the target vegetable at the prediction target time. From this, the effect of the present invention is clear.

<< Other Embodiments >>
Although the index value prediction device, the index value prediction method, and the program of the present invention have been described above with an example, embodiments other than the above-described embodiment can be considered. For example, the following variations can be considered for variables that are input values used for index value prediction and uses of the index value prediction device.

(Variation of variables)
In the above-described embodiment, in predicting the market price of the target vegetable at the forecast target time, the past weather of the specific area, the past market price of the non-target vegetable, and the past market price of the target vegetable are used as variables. However, variables other than the above may be added to predict the market price of the target vegetables at the forecast target time. That is, the computer constituting the price prediction device 10 has a fourth acquisition unit that acquires data corresponding to the additional variable, and each of the first acquisition unit, the second acquisition unit, the third acquisition unit, and the fourth acquisition unit. You may input the acquired data of the above into the prediction model to predict the market price of the target vegetable at the target time of prediction. The prediction model in this case, specifically, the prediction formula includes a term of additional variables indicated by the acquired data of the fourth acquisition unit and a parameter to be multiplied by the term.

Examples of additional variables include variables related to the economy and variables related to social conditions.
Examples of economic variables include the business conditions index of the country to which the target vegetable is produced or consumed, the gross domestic product (GDP), personal consumption, capital investment, housing investment, public investment, and exports. Import value, trade balance, production volume / shipment volume / inventory volume of various products, corporate profits, business conditions judgment index, number of bankruptcies, number of employment, prices, number of immigrants, etc. can be mentioned.
Variables related to social conditions include, for example, index values related to politics, social issues, public health, etc. of the country to which the target vegetable production area or consumption area belongs, specifically, infection with an infectious disease such as the new coronavirus. Daily or cumulative values for the number of persons, the number of severely ill and the number of deaths, as well as other indicators of domestic infection status.
Between the variables related to domestic infection status (specifically, the number of infected people this month / the number of infected people last month) and the market price (specifically, the monthly average wholesale price and total monthly wholesale price of a certain item of vegetables) , A clear correlation as shown in FIGS. 6A and 6B is confirmed. Therefore, by predicting the market price by adding a numerical value related to the domestic infection status as a variable, it is possible to predict the market price with higher accuracy, especially in a situation where the degree of infection of an infectious disease changes.

(Variations of device applications)
In the above-described embodiment, the index value predictor predicts the market price of the target vegetable at the prediction target time, and the producer of the target vegetable sets the production conditions of the target vegetable according to the prediction result, or the target vegetable It was decided to determine the sales line and adjust the shipping volume. However, the present invention is not limited to this, and the index value prediction device (specifically, the prediction unit 105) may predict the value related to the shipment scale of the target vegetable as the index value of the target vegetable. The value related to the shipping scale is the shipping amount or the shipping value, etc., and the value predicted by the index value prediction device is obtained by these future appropriate values, specifically, the producer of the target vegetable by shipping the target vegetable. It is a value that maximizes the profits that can be obtained.

When predicting the value related to the shipment scale of the target vegetable, for example, the market price of the target vegetable in the past each month is calculated by the above-mentioned procedure. In addition, the producer of the target vegetable is asked to input the past monthly shipment amount or monthly shipment amount of the target vegetable. Then, the correspondence (correlation) between the calculated market price of each month and the input monthly shipment amount or each month shipment amount is specified. After that, the market price of the target vegetables at the forecast target time is predicted, and the appropriate shipment amount or the appropriate shipment value of the target vegetables at the forecast target time is calculated (forecast) based on the predicted market value and the above correspondence relationship. The producer of the target vegetable can confirm the forecast result and determine the actual shipment amount and shipment value based on the forecast result.

10 Price predictor 11 Processor 12 Memory 13 Communication interface 14 Storage 15 Input device 16 Output device 101 1st acquisition unit 102 2nd acquisition unit 103 3rd acquisition unit 104 Prediction model storage unit 105 Prediction unit 106 Evaluation unit F Plant factory S Factory cultivation system

Claims (11)

It is an index value predictor that predicts the market index value of the target plant.
The first acquisition department that acquires data on past weather in a specific area,
A second acquisition unit that acquires data showing past market index values of non-target plants different from the target plants, and
A third acquisition unit that acquires data indicating past market index values of the target plant, and
A storage unit that stores a prediction model for predicting the market index value of the target plant,
The acquired data of the first acquisition unit, the acquisition data of the second acquisition unit, and the acquisition data of the third acquisition unit are input to the prediction model to predict the market index value of the target plant at the prediction target time. An index value prediction device having a prediction unit and a prediction unit. The second acquisition unit acquires data indicating past market index values of a plurality of types of the non-target plants, and obtains data.
Using the acquired data of the second acquisition unit, the calculation unit that calculates the market index value of the target plant at the past calculation target time while changing the type of the non-target plant,
The degree of correlation between the market index value of the target plant at the calculation target time indicated by the acquisition data of the third acquisition unit and the market index value of the target plant at the calculation target time calculated by the calculation unit is not determined. It has an evaluation unit that evaluates while changing the type of target plant,
By inputting the data acquired by the second acquisition unit to the prediction model for the non-target plant of the type selected based on the correlation degree, the prediction unit markets the target plant at the prediction target time. The index value prediction device according to claim 1, which predicts an index value. A plurality of combinations consisting of two or more types of the non-target plants are set.
The calculation unit uses the data acquired by the second acquisition unit for each of the two or more types of non-target plants included in each combination to obtain the market index value of the target plant at the calculation target time. Calculate while changing the combination of target plants,
The evaluation unit is the market index value of the target plant at the calculation target time indicated by the acquired data of the third acquisition unit and the market index value of the target plant at the calculation target time calculated by the calculation unit. The degree of correlation was evaluated while changing the combination of the non-target plants.
The prediction unit predicts by inputting data acquired by the second acquisition unit to the prediction model for each of the two or more types of non-target plants included in the combination selected based on the degree of correlation. The index value prediction device according to claim 2, which predicts the market index value of the target plant at the target time. The first acquisition unit acquires data on past weather in a plurality of the specific areas, and obtains data.
Using the acquired data of the first acquisition unit, the calculation unit that calculates the market index value of the target plant at the past calculation target time while changing the specific area, and the calculation unit.
The degree of correlation between the market index value of the target plant at the calculation target time indicated by the acquisition data of the third acquisition unit and the market index value of the target plant at the calculation target time calculated by the calculation unit is specified. It has an evaluation department that evaluates while changing regions,
By inputting the data acquired by the first acquisition unit for the specific area selected based on the degree of correlation into the prediction model, the prediction unit obtains the market index value of the target plant at the prediction target time. The index value prediction device according to any one of claims 1 to 3, which is used for prediction. Multiple combinations consisting of two or more of the specific areas are set,
The calculation unit uses the data acquired by the first acquisition unit for each of the two or more specific areas included in each combination to obtain the market index value of the target plant at the calculation target time in the specific area. Calculate while changing the combination,
The evaluation unit includes the market index value of the target plant at the calculation target time indicated by the acquired data of the third acquisition unit and the market index value of the target plant at the calculation target time calculated by the calculation unit. The degree of correlation is evaluated while changing the combination of the specific areas.
The prediction unit inputs the data acquired by the first acquisition unit for each of the two or more specific areas included in the combination selected based on the degree of correlation into the prediction model, thereby predicting the prediction target time. The index value prediction device according to claim 4, wherein the market index value of the target plant is predicted. The second acquisition unit acquires data indicating the market index value of the non-target plant in the period before the forecast target time.
The third acquisition unit acquires data indicating the market index value of the target plant in a period before the forecast target time.
The prediction unit
Among the acquired data of the second acquisition unit, data showing the market index value of the non-target plant at the specified time and
By inputting the data indicating the market index value of the target plant at the specified time out of the acquired data of the third acquisition unit into the prediction model, the market index value of the target plant at the prediction target time is input. The index value predicting device according to any one of claims 1 to 5, which predicts the above. The first acquisition unit acquires data on past weather in the specific area where the target plant is produced and at least one of the specific areas where the wholesale market for the target plant is located. The index value predicting device according to any one of 6 to 6. The index value prediction device according to any one of claims 1 to 7, wherein the prediction unit predicts a value related to the shipping scale of the target plant as the market price value of the target plant. It also has a 4th acquisition department to acquire data on the domestic infection status of infectious diseases.
The prediction unit inputs the acquisition data of the first acquisition unit, the acquisition data of the second acquisition unit, the acquisition data of the third acquisition unit, and the acquisition data of the fourth acquisition unit into the prediction model, respectively. The index value prediction device according to any one of claims 1 to 8, wherein the market index value of the target plant at the prediction target time is predicted. It is an index value prediction method that predicts the market index value of the target plant.
The first acquisition process, in which the computer acquires data on past weather in a specific area,
A second acquisition step in which the computer acquires data indicating past market index values of non-target plants different from the target plants.
A third acquisition step in which the computer acquires data indicating the past market index values of the target plant, and
A storage process in which a computer stores a prediction model for predicting the market index value of the target plant.
By inputting the acquired data in the first acquisition step, the acquired data in the second acquisition step, and the acquired data in the third acquisition step into the prediction model, the computer inputs the target at the prediction target time. An index value prediction method having a prediction process for predicting a market index value of a plant. A program that lets a computer predict the market index value of a target plant.
The first acquisition process to acquire data on past weather in a specific area,
The second acquisition step of acquiring data showing the past market index values of non-target plants different from the target plants, and
The third acquisition step of acquiring data indicating the past market index value of the target plant, and
A storage process for storing a prediction model for predicting the market index value of the target plant, and
By inputting the acquired data in the first acquisition step, the acquired data in the second acquisition step, and the acquired data in the third acquisition step into the prediction model, the market for the target plant at the prediction target time. Prediction process for predicting index values and
A program configured to let a computer perform.

Images